Seems like this is still failing to fix the test case in #863, need to think/dig a bit more.

Also commented on #863, but will duplicate here to keep info together:

Ok, I think I was overthinking/overdoing it. I've simplified #869 so that instead of trying to do escaping, it just splits based on &/; and :. That should get the original goal (normalizing so that if you use a hash to define the mock but a string to call it, or vice versa, it still matches). I think I was thinking about the real case (where escaping comes into play). Especially since this is all internal for the sake of matching stubs, the encoding wasn't really gaining us much even if it was working (and it wasn't working).